Land Reform and the PoliticalOrganization of Agriculture∗

Jonathan H. Conning� James A. Robinson�

April 2001

Abstract

The modern theory of agrarian organization has studied how the economic en-vironment determines organizational form under the assumption of stable propertyrights to land. The political economy literature has modelled the endogenous de-termination of property rights. In this paper we propose a model in which theeconomic organization of agriculture and the political equilibrium determining thedistribution of property rights are jointly determined. In particular, because theform of organization may affect the probability and distribution of beneÞts fromagrarian reform, it may be determined in anticipation of this impact. The modeloffers a reason for why tenancy, despite its economic advantages has been so littleused in countries where agrarian reform is a salient political issue. We argue thatthis in particular helps to understand the dearth of tenancy and the relative failureof land reform in Latin America.

Keywords: Agrarian Organization, Political Economy, Land Reform.

JEL ClassiÞcation: Q15, O12, N50, D72

∗We are grateful for the comments of seminar participants at the London School of Economics, TheWorld Bank, the NEUDC Conference at Cornell, and particularly to Tim Besley, Alain de Janvry, KlausDeininger, and Stephen Sheppard.

�Department of Economics, Williams College, Williamstown, MA 01267;e-mail: jconning@williams.edu.

�Department of Political Science, University of California, Berkeley, CA 94720;.e-mail: jamesar@socrates.berkeley.edu.

1

1 Introduction

The study of the nature and determinants of agrarian organization is one of the oldest

topics in economics and the system of metayage, or sharecropping, was discussed by

Adam Smith, J.S. Mill and Alfred Marshall. A dominant theme in the modern literature1

is that contractual arrangements and ownership patterns are determined by the incentive

problems that arise when labor effort or other relevant production actions are difficult to

observe, or costly to monitor. Theory tells us that a whole class of such incentive problems

can be solved, or at least ameliorated, by renting or selling the Þrm to the agent since

residual claimant status better aligns the agent�s objectives with those of the production

enterprise.

A successful theory of agrarian organization ought to be able to account for the large

differences across countries and continents, and evolution over time. For example, in the

absence of scale economies, this theory predicts that total land area under tenancy should

be higher in areas where land is more unequally distributed, as land lease markets ought

to reallocate land from relatively land abundant households. While actual comparisons of

aggregate tenancy patterns across regions are complicated by agro-climate and technolog-

ical considerations, we nonetheless should expect a relatively lower incidence of tenancy

in regions like Asia where land inequality has been historically low (particularly after mid

20th century land reforms in China, Japan, South Korea and Taiwan). Contrariwise, we

would expect a relatively high fraction of cultivated land to be organized under tenancy

in Latin America where land inequality has been extremely high.

That this is evidently not the case can be seen clearly in Tables 1 and 2. Table 1, which

is adapted from Hayami and Otsuka�s (1993; Table 1.1) survey of agricultural contracts,

indicates important differences in the use of tenancy across regions. In a sample of twelve

European countries for which comparable data was available over 40 percent of cultivated

land in 1970 was farmed under pure tenancy on land cultivated by owners who also leased

land. In the United States and Canada the comparable Þgure was over 60 percent.2 For1Cheung (1969) and Stiglitz (1974) are early examples of this approach. Literature surveys include

Bardhan (1989), Bardhan and Udry (1999), Basu (1997) and Hayami and Otsuka (1993).2These Þgures somewhat overestimate the actual extent of tenancy because they cannot distinguish

between owned and leased land operated by owner-cum-tenants. Recent Þgures from the 1997 US Agri-cultural Census help clarify this distinction however, by indicating that approximately 53 percent of landoperated by owner-cum-tenants was leased land. This leads to an estimate that approximately 49% ofharvested cropland in the United States was cultivated under leased land. Assuming the same ratio held

1

a sample of ten Asian countries, approximately 16 percent of land was under tenancy. As

Table 2 indicates however that prior to experiencing land reforms, countries such as Korea,

Japan and Taiwan had tenancy rates close to or above 50 percent. Latin America stands

out in sharp contrast to these other regions: despite having by far the most concentrated

land ownership pattern, and fewer laws to regulate tenancy contracts, less than 12 percent

of cultivated land was under tenancy in 1970.

Such differences across regions cannot be accounted for using existing microeconomic

theories except by assuming that fundamental differences in market structures or infor-

mation removed the advantages of tenancy in Latin America. Yet this is not self-evidently

plausible.3 Another potential explanation is that agro-climate or technological considera-

tions dictate that crops in Latin America are subject to greater scale economies. But this

too seems to fall short of a complete explanation for several reasons. First, the fraction of

cultivable land under plantation agriculture subject to technological economies of scale is

not large enough to explain the magnitude of the observed differences, except for certain

countries and regions (Sokoloff and Engerman, 2000). More importantly, comparisons of

agrarian organization on a crop by crop basis also reveal the relative paucity of tenancy

arrangements in Latin America compared to other parts. Finally, arguing that economies

of scale explains the difference seems to lead down the wrong path because the highest ex-

tent of tenancy is found in North America where heavy mechanization might be expected

to make economies of scale important.

In this paper we present a political economy theory of agrarian organization which can

help account for these facts. The modern literature has stressed the economic environment

as the key determinant of agrarian structure. In doing so, however, it has assumed

that property rights are secure. Yet, property rights over land have been contested and

redeÞned in almost all agrarian societies, and land reform has been one of the burning

political issues of the past century (Moore, 1965; Binswanger et. al, 1995). Although

land reform has at times taken place in revolutionary contexts, for example in Mexico,

China, or Cuba, a larger number of reforms have been implemented or attempted in the

context, or anticipation of, normal electoral competition. In Latin America important

for the data of Table 1, then approximately 40 percent of cultivated land in North America was undertenancy in 1970.

3Theories that appeal to credit market imperfections (e.g. Eswaran and Kotwal, 1986, Banerjee andNewman, 1993, Legros and Newman, 1996) or to the uneven distribution of non-traded skills may accountfor why Þxed rent tenancy may not be more prevalent amongst the poor in a given time period, and theseare surely important factors in Latin America. But these theories do not account for why the distributionof talent or Þnance should have remained so fundamentally different across regions over such long periodsof time (Carter and Zimmerman, 2000).

2

attempts at land reforms followed democratization in Bolivia, Chile, Colombia, Costa

Rica, Guatemala, Dominican Republic, and Venezuela (Lapp, 1997).4 Property rights are

the endogenous outcome of collective political choices and a striking difference between

Latin America and Western Europe or Asia in the 20th century, is the extent to which

property rights have been perceived to be stable and secure in the rural sector. These

facts suggest that agrarian structure could itself be affected, not just by the economic

environment, but also by the way in which property rights are determined and sustained.5

We propose a model in which the economic organization of agriculture and the political

equilibrium determining the distribution of property rights are jointly determined. We

identify a relationship between agrarian organization and the net beneÞts that agents can

expect via politically mediated land reform. SpeciÞcally, tenants may acquire skills, or

de facto property rights from squatting, which increase their potential gains under land

reform. This tends to increase its extent through the political system. In anticipation of

this, however, and despite possible economic beneÞts of tenancy, landlords may choose to

limit the extent of tenancy in order to increase the stability of their property rights. By

limiting tenancy landlords undermine squatting rights and slow skill accumulation in ways

that may limit the extent of land reform and the threat of expropriation. In section 3 we

document the importance of this mechanism for helping to account for the paradoxical

dearth of tenancy within Latin America�s landlord estates.

Nevertheless, as the conventional wisdom predicts and the Tables indicate, tenancy

is the prevalent institution in many parts of the world. Moreover, agrarian reforms have

taken place in situations where tenancy was once dominant, as the important examples

Japan, Korea and Taiwan illustrate, and tenancy reforms have been important through-

out Europe. Why should tenancy have been particularly threatening to landlords in Latin

America? Why did the forces that led to low tenancy in Latin America not operate in

East Asia and Europe? Our model suggests at least three reasons: (1) factor endow-

ments, pre-reform land inequality and competition; (2) higher basic education increases

the opportunity cost of restricting tenancy; and (3) how land reform emerges as a national

political issue.

Table 2 shows pre-land reform land concentration and tenancy for selected countries

for which data is available. The Þrst comparative static result of our model suggests that4Political reforms that extended voting rights to tenants and small farmers also led to signiÞcant

changes in tenancy regulation and land taxation in a large number of European countries (Swinnen,2000).

5By this we do not just mean the well known effects of the instability of property rights on investment(Besley, 1995).

3

in countries where per capita land inequality is highest, there is a greater incentive to

challenge property rights via the political system, and this makes landlords more likely to

organize agriculture in a politically defensive manner, by limiting tenancy. This suggests

that one reason why tenancy is less threatening to landlords in Western Europe or North

America is that land is much more equally distributed. All else equal, the model also

suggests that Latin America�s historically higher land-to-labor ratio has also acted to

increase the political risk of land reform.

Result (2) suggests that differential human capital attainment might also help explain

the relative incidence of tenancy. In our model human capital may raise the economic

beneÞts of tenancy relative to its political cost, suggesting another reason why we are

more likely to observe tenancy in the Asian context, or in countries such as Argentina

and Uruguay which have long lead the continent in educational achievement.

How land reform emerges as a national political issue may also determine landlords�

production organization decisions. Within Latin America, in situations where land reform

has not become an important political issue, we ought to see a greater extent of tenancy.

This appears to be the case in countries like Uruguay, Argentina, and Chile. Uruguay

stands out for having both a very high land Gini, and the highest extent of tenancy of

the six countries listed in Table 2. Argentina has also been noted for its relatively high

tenancy rates (Taylor, 1997). Yet in both of these countries, political cleavages have

never coalesced around the issue of agrarian reform, but instead around the urban-rural

dichotomy. As a result, land reform has not been a national political issue because the

political parties were more interested in redistributing between the urban and rural sectors

rather than within the latter.6 Chile is an interesting intermediate case where an urban-

rural cleavage temporarily gave way to a rich-poor cleavage after 1958 and a short-lived

opening for land reform.

Another important instance where result (3) sheds light on agrarian structure is in

East Asia. Land reforms that affected Taiwan, Korea and Japan, each occurred in the

context of actual or threatened external invasion which were unanticipated events that

undermined previously dominant landlord classes whose authority had previously been

unchallenged. However, as soon as the political basis for these property rights was un-

dermined, the existence of a large population of tenants assured strong political support

for far reaching land reforms, as the case of Korea described below clearly attests. The6This seems to have been the case both in Argentina where the urban Peronist coalition redistributed

to itself away from the rural sector (see O�Donnell, 1979). Similarly, in Uruguay, the split was betweenthe urban Colorados and the rural Blancos (Collier and Collier, 1991, Chapter 5).

4

fact that land reform arose in an unanticipated manner meant that landlords had not

organized production to avoid this, and this led to large reforms since tenants were able

to effectively use redistributed land. In Latin America reforms were more often de-

railed, limited or reversed by the politically defensive actions of landlords. Furthermore,

were land reforms did take place the disorganization costs were often high as land was

transferred to households with more limited farm management experience.

Our paper is related to several literatures. The connection between the anticipation of

land reform and defensive patterns of agrarian organization has been discussed and amply

documented by many scholars including Palacios (1979), de Janvry (1981), Le Grand

(1986), Zamosc (1986), and Binswanger et. al. (1995) amongst others. Nevertheless

none of these studies provides an explicit mechanism linking the two issues, nor do they

fully explain in comparative perspective why land reform should become a political issue

in some countries rather than others or why agrarian organization differs so much even

where land reform had become a salient political issue. This is a major contribution of

our analysis and a key for understanding the relative efficiency of land reforms.

As we noted above, our paper is the Þrst to depart from the standard approach to

agrarian structure by emphasizing the role played by the political, rather than the eco-

nomic, environment. Several models, for example Grossman (1993), Horowitz (1993),

and Acemoglu and Robinson (1999), have examined the incentive to redistribute land

as a way of forestalling social conßict or revolution.7 Our model differs in focusing on

non-revolutionary politics and studying the joint determination of land reform and the

organization of production. This allows us to derive new comparative results regarding

the incidence and success of reform. Finally, our research is related to a large political

economy literature which has stressed how inefficient decisions may arise to manipulate

future political equilibria. This research includes Persson and Svensson (1989), Alesina

and Tabellini (1990), Aghion and Bolton (1990), and Besley and Coate (1998) in the

context of democratic politics, and Robinson (1998) and Bourguignon and Verdier (2000)

in non-democratic polities. Apart from the different focus and motivation of our analysis,

the fundamental theoretical difference is that in our model it is private agents and not

political decisionmakers who take actions that affect subsequent political outcomes.

The rest of the paper is organized as follows. The next section presents the basic

model and analyses comparative static results of the political-economic equilibrium that7This view has evidently also inßuenced policymakers and military strategists who have often placed

land reform at the center of counterinsurgency plans in countries from Vietnam to El Salvador (Proster-man and Reidinger, 1987; Prosterman et al., 1990).

5

it implies. Section 3 discusses a number of historical episodes that appear to be consistent

with the interpretations given here and Section 4 concludes.

2 Elements of a Model

2.1 Fundamentals

We consider a two period economy. In the Þrst period the distribution of land is given and

the organization of agriculture is determined for both periods. At the start of the second

period there is an election where two office motivated political parties compete for power

by offering to redistribute land.8 The outcome of this electoral competition determines

the equilibrium amount of land reform. After the election, land reform is implemented

and second period production takes place.

There are three types of agents. In the rural sector there is a mass λL of �landlords�

and a mass λP are �peasants.� The urban sector has a mass of agents of measure λU . We

adopt the normalization that λL+ λP + λU = 1. Each agent tries to maximize the sum of

consumption over the two periods and there is no discounting. All agents also gain utility

from whichever political party is in power (their �ideological preferences�) which will be

important subsequently in determining the outcome of the election.

We shall model the urban sector in a reduced form way and assume that all agents

there have an exogenous endowment income of y. All production and resource allocation

therefore takes place in the rural sector. In the rural sector of the economy, there is an

endowment of T units of land initially divided between landlords and peasants. Each

agent in the rural economy is also exogenously endowed with h units of human capital.

We assume that θ is the proportion of total land owned by the landlords so that each

landlord holds θT/λL while each peasant has (1 − θ)T/λP units of land. Naturally, werequire that landlord households own more land than peasants,

θT

λL>(1− θ)TλP

=⇒ θ >λL

λL + λP.

There is a single consumption good which is numeraire with price normalized to unity.

The good is produced in the rural sector using capital and labor according to the constant

returns to scale production function, F (T, L) = s(t, h) (T + L). Here s(t, h) is total factor

productivity which in turn depends on two accumulated factors. Firstly, it depends on8Our political model is closely related to the probabilistic voting model of Lindbeck and Weibull

(1987), Dixit and Londregan (1996) and Persson and Tabellini (2000).

6

peasant skill which is in turn assumed to rise with average farming experience and hence

with t ∈ [0, 1], the amount of land that a landlord places under tenancy in the previousperiod. To keep things simple we assume that t represents the fraction of land worked as

a tenant by each peasant so that all peasants are identical and there is no heterogeneity

between peasants who were or were not tenants in the Þrst period. Secondly, h is the

exogenous stock of human capital of each worker. We assume that sh > 0, shh < 0,

st > 0, stt < 0 and that tenancy and human capital are Edgeworth complements so

that sth > 0. The assumption of a linear production captures the idea that tenancy has

beneÞcial economic effects in the simplest possible way.9

By determining the extent of tenancy in the Þrst period landlords affect skill accumu-

lation, and hence productivity in the second period. For convenience only, we normalize

Þrst period t to zero and write the productivity level s(0, h).

2.2 Political institutions

Potential asset redistributions are mediated through the political process. We shall model

society�s choice regarding the possibility and extent of land reform as being the outcome

of electoral competition. To Þx ideas, we Þrst analyze the case of electoral competition

within the context of a probabilistic voting model, but then explain how the essential

trade-offs of the model would be adapted to other models of political competition.

Assume that there is one vote per-agent in the economy. Two political parties, which

we denote A and B compete for these votes. In this section we assume that both parties

have the sole objective of maximizing the probability of winning the election. We assume

that there is a single policy issue or instrument which is the proportion of land, denoted

α ∈ [0, 1], to be taken from the landlords and divided equally between the peasants. We

assume, however, that it is costly to redistribute land. To redistribute a proportion α

clearly costs landlords in terms of lost assets and reduced income but we also assume that

land reform imposes cost C(α) on each landlord, peasant and urban agent. Here C is

strictly increasing and convex with C(0) = 0, C 0 > 0, C 00 > 0 and C 000 ≤ 0.10 Clearly,9In an earlier draft we modeled farming skill s(t, h) as a non-traded production factor in a more general

constant or diminishing returns to scale production function F (T,L, s(t, h)). Suppressing tenancy thennot only slowed skill accumulation as in the linear model, but also, depending on the assumed marketstructure, could produce within period allocative inefficiencies. The linear model was adopted to vastlysimplify the comparative static analysis while retaining the essential insights.10The assumption that urban agents incur costs from land reform captures the idea that the costs of

agrarian reform spill over into the cities. This could be because the government raises taxes to Þnancereforms, because reform induces higher food prices, or perhaps because focusing policy on the rural sectorhas opportunity costs in terms of spending in the urban sector.

7

since neither landlords nor urban voters beneÞt from the redistribution of land, but both

bear costs, they strictly prefer α = 0.11 Peasants however potentially prefer α > 0. The

equilibrium extent of land reform therefore depends on how the political system aggregates

the preferences of different agents.

Let V g(αx) be the indirect utility of an agent of group g ∈ {L,P, U} as a functionof the extent of land reform offered by party x ∈ {A,B}. We assume that each agentreceives an aggregate ideological shock in favor of party B of δ and also an individual

speciÞc shock of σig. Thus an agent of group g votes for party A if the indirect utility

he gets from the policy platform of party A is greater than the indirect utility from the

policy of party B plus the ideological shocks. This implies,

V g(αA) > Vg(αB) + σ

ig + δ.

We assume that both σig is distributed uniformly on the intervalh− 12φg, 12φg

iand δ on the

intervalh− 12ψ, 12ψ

i. We can therefore calculate the critical value of the idiosyncratic shock

which leaves an agent indifferent between the parties. This is,

bσig = V g(αA)− V g(αB)− δ.All agents of group g with σig ≤ bσig vote for party A. The total number of agents ingroup g that vote for party A is therefore,

λgZ V g(αA)−V g(αB)−δ

− 12φg

1di = λgÃV g(αA)− V g(αB)− δ + 1

2φg

!

This follows because, for given δ, A gets the votes of all the agents of a group who have

low values of σig. The probability that party A wins the election, denoted χ(αA,αB), is

therefore the probability that the total number of votes it gets is at least one half of the

population, or,

χ(αA,αB) = Pr

(Xg

λgÃV g(αA)− V g(αB)− δ + 1

2φg

!≥ 1

2

)

which is,

Pr

(Xg

λgφgδ ≤Xg

λgφg (V g(αA)− V g(αB)))

11We are ruling out the case of α < 0 or �land grabs� where landlords seize peasant land, although thereis nothing in principle to rule this out. For example, late19th century Liberal reforms in several LatinAmerican republics privatized communal lands leading in several cases to large scale transfers of land tolandlords.

8

Integrating out over the support of δ we Þnd,

χ(αA,αB) =1

2+φ

ψ

Xg

λgφg (V g(αA)− V g(αB)) , (1)

where φ =Pλgφg. The probability of winning is a simple weighted sum of the utility

differences that the policy platforms of the parties induce.

A pure strategy Nash equilibrium between the parties is a pair of platforms (eαA, eαB),such that eαA = arg max

αA∈[0,1]χ(αA, eαB),

while eαB = arg maxαB∈[0,1]

1− χ(eαA,αB).By symmetry, a Nash equilibrium between the parties involves eαA = eαB = eα, where eαsatisÞes the Þrst-order condition,

Xg

ηg∂V g(eα)∂α

= 0. (2)

where ηg = λgφg. This Þrst order condition can be easily seen to be a weighted average of

the Þrst-order conditions that determine the preferred policy of each group. The weight

ηg given to each group�s preferences is proportional to that group�s population share of the

vote but is also affected by φg, which is the density of the ideological shocks σig. The larger

is φg, the narrower is the rangeh− 12φg, 12φg

iand hence the more ideologically homogenous

is group g. Groups with relatively larger φg will have more inßuence in determining the

equilibrium policy put forward by political parties because such groups will contain more

�swing voters,� or voters who are less likely to be individualistically swayed by ideology

or other superÞcial candidate characteristics, and more likely to vote in line with their

common economic interests. Thus candidates tilt their platforms toward the policies

desired by these groups.

Many other models of political competition boil down to an equilibrium condition such

as (2). For example, the simplest median voter model assigns ηg = 1 to the median voter

group and ηg = 0 to all other groups. As discussed below interesting equilibria with a

positive probability of land reform will arise within the context of this simple probabilistic

voting model only when the peasant group can achieve sufficient �political clout� to attract

the attention of political candidates. More generally, land reform will emerge within any

political system that gives sufficient political weight ηP to the peasant sector�s preferences.

9

2.3 Structure of Payoffs

The key interaction in the model is the effect of the organization of agriculture on the

extent of land reform determined by the political equilibrium. To keep things as simple

as possible, the fraction of land kept under tenancy in the Þrst period will not affect

current output, but does affect total factor productivity in the second period. Since

st > 0 it is clearly socially efficient to set t = 1. However, as we shall shortly show,

this is typically not an equilibrium because tenancy, by increasing the productivity of

land in the second period, may simultaneously undermine landlords� property rights by

increasing the attractiveness of land reform. Thus the higher is t the greater the extent

of land redistribution that the political parties will offer. This reduces the incentive to

landlords to set high t.

Landlord production organization decisions may be affected by the anticipation of re-

form, whether or not landlords internalize how their actions might inßuence the political

equilibrium. For example individual landlords might limit the extent of tenancy for fear of

a land-to-the-tiller land reform and the aggregate effect of their uncoordinated decisions

will affect the probability or extent of such reforms. In this paper we assume landlords do

internalize how their actions to try to affect the future political equilibrium. When this is

the case it seems reasonable to also assume that landlords understand how their actions

inßuence market prices. Under the assumed linear production technology this assumption

is without consequence however because current period marginal factor products are un-

affected by current production decisions. Hence factor prices will be the same whether

landlords are assumed to collude or not.12 Thus, though we describe landlords as being

able to collude, the analysis that follows could just as well have been described by the

assumption of competitive factor markets.

Since each peasant household has (1− θ)T/λP units of land, a peasant household canproduce income of

F (T, 1, s(0, h)) = s(0, h)

Ã(1− θ)TλP

+ 1

!(3)

by simply withdrawing from markets and reverting to autarchy. Thus landlords can never

drive them below this payoff. When landlords can make take-it-or-leave-it offers, (3) will

be the income and utility level of a peasant household in the Þrst period.

In the second period, however, peasants� income will increase, both due to the impact

of Þrst period tenancy, and also because as a result of land redistribution each household12As discussed in the next footnote, with a more general production technology, market structure does

matter for determining relative factor prices and income.

10

will have more land. In the second period peasant income level is

s(t, h)

Ã(1− θ)T + αθT

λP+ 1

!(4)

Landlord income in each period can be expressed as the entire output of the economy,

minus what they have to pay to satisfy the peasant sector�s reservation utility level. Thus

Þrst period landlord proÞts, denoted π1(θ, h), are;

π1(θ, h) = s(0, h)³T + λL + λP

´− λP s(0, h)

Ã(1− θ)TλP

+ 1

!

while second period proÞts, denoted π2(t,α, θ, h) are,

π2(t,α, θ, h) = s(t, h)³T + λL + λP

´− λP s(t, h)

Ã(1− θ)T + αθT

λP+ 1

!(5)

Note that, because of constant marginal products, Þrst period proÞts, π1(θ, h) is indepen-

dent of t. Altering the extent of tenancy in the Þrst period has no effect on Þrst-period

production because total factor productivity is predetermined in that period.13 However,

given α, π2(t,α, θ, h) is increasing in t.

2.4 Timing of the Game

We shall now calculate the pure strategy subgame perfect equilibrium of this game.

� Landlords determine the extent of tenancy.

� The political parties simultaneously and non-cooperatively determine their plat-forms.

� First period production, consumption and voting take place.

� The outcome of the election is determined and land reform is implemented.14

� Second period production and consumption take place.13With a more general production technology, Þrst-period allocative efficiency would be affected since

the suppression of tenancy would in equilibrium lead landlord farms to adopt higher land-to-labor ratioscompared to peasant farms. The size of this distortionary effect will in general depend on the assumedtechnology and market structure. Conning (2001) analyzes the consequences of tenancy suppression in astatic general equilibrium model without politics.14Since the political parties only care about the probability of winning and not the actual policy they

adopt it is weakly optimal for them to actually choose the α they promised. We therefore abstract fromissues of commitment to policy now, but relax this assumption later.

11

2.5 Analysis

Starting in the second period, we can now use (2) to calculate the equilibrium policy

adopted by the political parties for a given level of tenancy t and other parameters,

denoted α(t, θ, h, ), where η = (ηL, ηP , ηU). Substituting the payoffs into (2) we Þnd;

ηPÃs(t, h)

θT

λP− C 0(α)

!− ηL

Ãs(t, h)

θT

λL+ C 0(α)

!− ηUC 0(α) = 0

or, simplifying ³φP − φL

´s(t, h)θT =

Xg

ηgC 0(α) (6)

Equation (6) shows that the equilibrium amount of land reform, α(t, θ, h,η), is a weighted

function of the marginal effects of land reform on the utilities of the different voters. Note

that by the convexity of C, the second order condition is satisÞed.

For there to be any land reform α > 0 in equilibrium it must be the case that φP > φL.

This condition requires that the peasant sector have sufficiently more political clout than

the landlord sector, a condition that is not always likely to be met, even though peasants

are a much larger portion of the electorate.15 In what follows we assume that this

condition is met. Under the simplest median voter model with λP ≥ 12, ηP = 1 and

ηL = ηU = 0 the maximal extent of land reform would result and be determined by

s(t, h)θT = C 0(α).

The comparative statics of (6) are critical for our analysis. Note Þrst that,

∂α

∂t=(φP − φL)st(t, h)θTP

g ηgC 00(α)> 0. (7)

The higher is t the greater the productivity of land in peasant hands and hence the

higher the marginal beneÞt to a peasant from land reform. For a Þxed political weight,

this greater intensity of preference for reform induces the parties to offer more land redis-

tribution in their platforms. Thus higher tenancy leads to a higher proportion of landlord

land being redistributed. Straightforward calculations also give the following comparative

statics,∂α

∂θ> 0;

∂α

∂h> 0;

∂α

∂T> 0; and

∂α

∂φU< 0.

The greater is the initial inequality of land, the more a peasant beneÞts from a given

extent of land reform, and the more reform he wants. This induces more land reform in15Land reform has often emerged as key electoral issue only following political mobilization of peasant

groups by partisan political parties or outside groups (Kauffman, 1972; Lapp, 1997; Tuma, 1965).

12

the resulting political equilibrium. Similarly, the greater is the average stock of human

capital, the greater is productivity and the greater the preferred α. Again, this moves

the political equilibrium towards more reform. Also, the higher the land to labor ratio

(captured by T ) the greater the per-capita beneÞt from reform for peasants and hence the

desired α. This similarly increases equilibrium α(t, θ, h,η). Finally note that the greater

is φU the greater the political weight given to their preferences. Since urban voters prefer

α = 0 this reduces the extent of land reform offered by the two parties.

For later use it is also useful to consider the effects of changes in θ and h on the

relationship between tenancy and the extent of land reform - on ∂α∂t. To investigate this

we can calculate the following cross-partial derivatives (suppressing the arguments of the

functions),∂2α

∂t∂θ=(φP − φL)stTP

g ηgC 00− C

000

C 00∂α

∂t

∂α

∂θ> 0, (8)

∂2α

∂t∂h=(φP − φL)sthθTP

g ηgC 00− C

000

C 00∂α

∂t

∂α

∂h> 0. (9)

The signs follow from the previous comparative static results and the assumptions that

sth > 0 and C 000 ≤ 0. Expressions (8) and (9) are both intuitive. The Þrst says that thegreater is θ the greater is the marginal effect of tenancy on the extent of land reform. (9)

tells us that an increase in the level of human capital has a similar effects. The important

implication of these derivatives is that they show that when either initial land inequality is

higher, or the labor force is more educated, land reform becomes more sensitive to tenancy

decisions and the organization of agriculture. As we shall see, this tends to increase the

readiness of landlords to alter organization to affect the political equilibrium.

Having solved for both the political determinants of land reform in the second period

and the allocation of resources, it remains to determine the equilibrium amount of tenancy.

t is chosen by the landlords to solve the problem,

maxt∈[0,1]

π1(θ, h) + π2(t,α(t, θ, h,η), θ, h).

By construction, we know that Þrst-period proÞts are independent of t thus the Þrst-order

condition for this program simply maximizes (5), giving,

st(t, h)h(1− α(t, θ, h,η)) θT + λL

i− s(t, h)θT ∂α(t, θ, h,η)

∂t≤ 0. (10)

The Þrst-order condition clearly shows the trade-off between the beneÞcial effects of an

increase of tenancy on productivity that the landlord can capture, versus the greater

property rights insecurity effect of higher t increasing α. We assume the second-order

13

condition is satisÞed and denote the interior solution to this equation t(θ, h,η). The

implications of (10) are summarized by the following result.

Proposition 1 If the cost of redistributing land is sufficiently low, then the anticipation

of the impact of agrarian organization on the extent of land reform leads to an inefficiently

low level of tenancy so that t(θ, h,η) < 1.

Tenancy is reduced below the socially efficient level in an attempt to reduce the extent

to which agrarian reform is adopted as a policy by the political parties.16 Before consid-

ering the further implications of (10) and Proposition 1 it is important to think about

how our main result generalizes to a higher dimensional policy space. So far we have only

allowed for a single redistributive instrument - land reform. Why not also consider income

redistribution, particularly if land reform has the potential inefficiencies we isolate? In

our model, since there is no direct cost to landlords, tenancy would not be deterred by the

prospect of income redistribution in the second period. There are several reasons however

why, despite the availability of income redistribution, we would expect land reform to be

a key redistributive tool (as the evidence suggests that it indeed has been, particularly in

Latin America).

Firstly, few Latin American countries had the bureaucratic capacity to levy regular

income taxes until recently. Thus a once and for all redistribution of assets, despite

its costs, might generate a greater present value income for peasants. Secondly, income

taxation does have costs, the present value of which has to be compared to cost of land

reform. Finally, an important difference between land and income is that it is hard to

hide land while income can be hidden and moved abroad, making it very hard to tax.

Landlords might be better off if they could commit to pay income tax since this might

potentially remove the incentive of the peasants to support land redistribution, but they

may not be able to commit ex post not to hide their income.

Condition (10) however leaves open the question of why there is tenancy anywhere.

Why would this effect be more important in Latin America? To attack this question

we consider the comparative statics of (10). Our Þrst result is that under plausible

assumptions, greater initial land inequality lowers the extent of tenancy, ceteris paribus.

16A sufficient condition for this proposition is that st(1, h)h(1− α(1, θ, h,η)) θT + λL

i−

s(1, h)θT ∂α(1,θ,h,η)∂t < 0. This would be satisÞed if ∂α(1,θ,h,η)∂t is sufficiently large which will be true

if C00(α(1, θ, h,η)) is small.

14

To see this one implicitly differentiates (10), to obtain

dt

dθ=sth(1− α)− ∂α

∂θθiT − sT

h∂α∂t+ θ ∂

2α∂t∂θ

iΦ

(11)

where Φ = −stth(1− α) θT + λL

i+ 2st

∂α∂tθT + sθT ∂

2α∂t2

> 0 from the second-order condi-

tion.

Expression (11) shows that when initial land inequality is higher there are several

effects. First, the greater is θ, the greater are the beneÞts from the increased total

factor productivity that accrue to the landlords, and hence the inclination to choose a

higher t - this is the term st(1 − α)T . On the other hand, three other effects push to

make tenancy increasingly costly for landlords. Firstly, for any level of t, greater land

inequality increases the extent of reform (∂α∂θ> 0) and increases the sensitivity of reform

to t ( ∂2α

∂t∂θ> 0). Finally, higher θ gives landlords more to lose from reform, and this also

induces lower t since ∂α∂t> 0. Under our assumptions however, the net effect is that:

Proposition 2 The greater is initial land inequality, the lower is the incidence of ten-

ancy.

To see this notice that (11) above can be rearranged to obtain

dt

dθ=

hst(1− α)− s∂α∂t

iT −

hst∂α∂θ− s ∂2α

∂t∂θ

iθT

Φ(12)

=−hstλ

L/θi−hst∂α∂θ+ s ∂

2α∂t∂θ

iθT

Φ< 0 (13)

In the last step we�ve used −hstλ

L/θi=hst(1− α)− s∂α∂t

iT which follows from (10)

above. From earlier assumptions and results ∂α∂θ> 0, ∂2α

∂t∂θ> 0, Φ > 0, which lead

to dtdθ< 0.

This monotonic relationship results under our assumed constant returns to scale pro-

duction technology because under secure property rights full tenancy is the efficient out-

come regardless of initial inequality. Increasing property rights insecurity can therefore

only lead landlords to reduce tenancy relative to this efficient benchmark. Under a

decreasing returns to scale production technology we might instead expect an inverse

U-shaped relation between initial inequality and tenancy. This is because under secure

property rights the tenancy should rise with initial inequality as the lease market reallo-

cates land to reach the efficient determinate pattern of operational farm sizes. At low

to medium levels of land concentration, these allocative efficiency considerations would

15

dominate so tenancy would rise with inequality, but at higher levels of concentration

the property insecurity effect would rise in importance leading to an emergent inverse

relationship between tenancy and higher initial inequality.

This is our Þrst explanation of why the incidence of tenancy is so low in Latin America.

The model suggests another explanation however. To see this we now investigate the effect

of human capital h on the optimal choice of t,

dt

dh=sth

h(1− α) θT + λL

i− st ∂α∂hθT − θT

hsh∂α∂t+ s ∂

2α∂t∂h

iΦ

(14)

Expression (14) shows how differences in the educational attainment of the labor force

inßuence landlords� optimal trade-offs. Again there are offsetting forces. On the one

hand, when peasants become better educated, they can use land more productively and

this increases the amount of land reform they prefer. These preferences induce the political

parties to adopt platforms offering more reform (∂α∂h> 0). Moreover, when h is higher,

α is higher for any given level of tenancy ( ∂2α

∂t∂h> 0). A further effect leading to lower

tenancy stems from sh > 0 which increases the marginal loss to the landlords from reform.

However, the Þrst-term in numerator of (14) pushes towards higher t. This is because if

human capital h and farming skill from tenancy t are complementary, the higher is the

human capital level of peasants, the greater the opportunity cost of reducing tenancy (in

terms of foregone land rents), and hence other things equal, the greater is t. Therefore, if

sth is sufficiently strong then we can have the following result.

Proposition 3 If human capital and the productivity gains from tenancy are sufficiently

complementary, then the higher is the human capital attainment of the labor force, the

greater the extent of tenancy.

Proposition 3 provides further clues as to what might be different about Latin America.

Not only has initial land inequality been much greater than in either Europe, North

America or Asia, but rural educational attainment has also lagged considerably. Both

factors may lead the political advantages of reducing tenancy to dominate the productivity

enhancing ones. Such differences in impact may also help account for elites� differential

willingness to invest in public education across these different regions.17

Other factors may determine the extent to which land reform is an important national

political issue. To the extent that it is not, then there is little incentive for landlords to17Bourguinon and Verdier (2000) discuss elites� reluctance to invest in public education for fear that

education will increase the size of the voting population, and Sokoloff and Engerman (2000) document theempirical relationship in Latin America between land inequality, democratization, and public education.

16

organize production defensively. An obvious reason why land reform may not become a

salient political issue would be that the majority of voters were urban, or more generally

that urban voters exert great political power. Intuitively, the greater is the political weight

of the urban sector, the less interested are the majority of voters in land reform and the

less important it becomes necessary to attract the rural vote. This tends to increase

the equilibrium degree of tenancy. We can show this result formally in our model by

investigating the effects of a change in φU .

dt

dφU=−st ∂α∂φU θT − sT θ ∂2α

∂t∂φU

Φ> 0 (15)

In (15) the Þrst term.18 We therefore have the following;

Proposition 4 The greater is the political weight of the urban sector, the greater is the

equilibrium level of tenancy.

2.6 Extension: Political Cleavages and Agrarian Reform

We now develop a simple extension to capture the idea that the nature of political cleav-

ages may be an important determinant of the extent of land reform offered by political

parties and therefore of the pattern of agrarian organization. As we noted in the intro-

duction, those Latin American countries where we observe a large incidence of tenancy

are also those where redistributive politics evolved into a battle, not between the rich and

the poor, but rather between urban and rural constituencies. We show that if this is the

case then it is likely that the equilibrium extent of land reform will be lower and hence

the incidence of tenancy higher.

There are now two types of urban agents, the poor of mass λUP and the rich of mass

λUR. The rich have a higher exogenous income endowment of yR and yP , but since

both groups only bear the costs of land reform and get no beneÞts they prefer α = 0.

Intuitively, when there is a rich versus poor cleavage, peasants and the urban poor form a

coalition. The urban poor do not desire land reform and this tends to reduce α below the

level desired by the peasants. However, when there is an urban versus rural cleavage, the

peasants are in a coalition with the landlords. Since landlords plausibly dislike land reform

much more than the urban poor this leads to a lower equilibrium level of α. This scenario

can help explain the prevalence of tenancy in countries like Argentina and Uruguay.19

18Note, ∂2α∂t∂φU

= −∂α∂t

µλUPgηg+ C000

C00∂α∂φU

¶< 0.

19The case of Chile, another early urbanized economy, is discussed in the next section.

17

To model how cleavages affect the equilibrium level of land reform we extend the model

of the previous section not only by introducing the new urban group, but by also allowing

the political parties to have policy preferences. Rather than simply being interested in

winning power, parties now care directly about the policy they implement. Let Wx(α)

be the utility function of party x. We consider two different scenarios and compare the

expected extent of land reform in a political equilibrium. In the case of an urban versus

rural cleavage, party A�s utility is a function of the utilities of landlords and peasants,

while party B�s utility is a function of urban poor and rich. Let these be respectively,

WA(α) = ρPAVP (α) + (1− ρPA)V L(α)

WB(α) = ρURB V UR(α) + (1− ρURB )V UP (α)

where ρPA and ρURB are the weights given to the peasants and the urban rich by parties A

and B respectively.

V L(α) = s(t, h)³T + λL + λP

´− λP s(t, h)

Ã(1− θ)T + αθT

λP+ 1

!− C(α)

V P (α) = s(t, h)

Ã(1− θ)T + αθT

λP+ 1

!− C(α)

V UR(α) = yR − C(α)V UP (α) = yP − C(α)

We can calculate the preferred policies of the two parties by maximizing these functions.

Clearly, argmaxαWB(α) = 0, and let αA = argmaxαWA(α).

In the poor versus rich cleavage scenario we instead assume that party A represents

the peasants and the urban poor, while party B represents the landlords and the urban

rich. In this case their respective party welfare functions are

WA(α) = ρPAV P (α) + (1− ρP

A)V UP (α)

WB(α) = ρURBV UR(α) + (1− ρUR

B)V L(α)

where ρPAand ρUR

Bare again the weights attached to different groups� welfare within the

parties. Clearly, argmaxαWB(α) = 0, and let αA = argmaxαWA(α).

The timing of the game is as before. However, in the case where parties have policy

preferences the issue of commitment becomes important. We conduct the analysis by

assuming, as in Alesina (1988) and Besley and Coate (1997) that parties cannot make

commitments. This implies that whichever party is elected will then simply choose its

18

ideal point as the policy. Anticipating this, voters vote for whichever party�s ideal point

is closest to their own preferred policy. With these assumptions it is easy to calculate the

probabilities that either type of party A wins the election. These are,

χ(αA, 0) =1

2+Xg

λg (V g(αA)− V g(0)) . (16)

χ(αA, 0) =1

2+Xg

λg (V g(αA)− V g(0)) . (17)

Thus the expected extent of land reform in these two scenario cases is, χ(αA, 0)αA and

χ(αA, 0)αA, respectively.

Taking the case where ρPA= ρPA it is immediate αA > αA. This follows simply from

checking the Þrst-order conditions for the maximization of WA(α) andWA(α). With this

we can now state the main result of this section,

Proposition 5 Conditional on observing land reform, the extent of land reform is greater

when politics are dominated by poor vs. rich cleavages than when they are dominated

by rural vs. urban cleavages. Moreover, if λP is sufficiently high, then χ(αA, 0)αA <

χ(αA, 0)αA, so that in addition the expected equilibrium extent of land reform is higher.

Intuitively, although the urban poor dislike land reform they do so much less than

landlords and so when party A forms along class lines it prefers a greater degree of land

reform. However, it does not follow from this that the expected amount of land reform

is higher when αA > αA. This is because a party representing the poor may offering a

platform which is further from the platform offered by the party representing the rich,

than the policy offered by a rural party is from that offered by the urban party. Thus

it could be that the pro-poor party gets elected less often than a pro-rural party, i.e.

χ(αA, 0) < χ(αA, 0). The condition on the size of λP takes care of this case since clearly

the poor prefer αA to αA so that λP (V P (αA) − V P (αA)) > 0, which tends to increase

χ(αA, 0) relative to χ(αA, 0). Thus, if the mass of peasants is large enough χ(αA, 0)

> χ(αA, 0) which is sufficient (though not necessary) for the expected extent of land

reform to be higher when there is a poor vs. rich cleavage..

3 Discussion and evidence

A large number of historical studies illustrate the use of defensive patterns of agrarian

organization to protect against the real or perceived threats of property rights challenges.

19

In some instances the connection between the form of agrarian organization and the

protection of property rights is patently obvious. For example in El Salvador in the

early eighties thousands of tenants were evicted shortly after it became apparent that

legislation for a land-to-the-tiller agrarian reform was being discussed (Pelupussy, 1996;

Prosterman and Riedinger, 1987). In other instances however, the connection may be

much less apparent, perhaps because the property rights conßicts are latent. Barrington

Moore (1965) and Scott (1985) are just two classics in a large political and anthropological

literature that explores the actual or latent nature of property rights conßicts over land,

and their role in shaping political and economic outcomes.

de Janvry (1981) is a classic statement of the ways that land rental and sales markets

have failed to reallocate land toward family farmers in Latin America, and how land reform

processes have been subverted or stopped through the political activities of landlords .

He argues that in several countries the anticipation of land reform landlords expelled

tenants and turned to mechanization and new crops which relied on hired wage labor

rather than tenants (de Janvry and Sadoulet, 1993). Binswanger, Dieninger, and Feder

(1995) summarize a large body of historical evidence in support both of the stylized fact

that farms which rely primarily on family labor enjoy a productivity advantage over wage-

labor operated farms, and yet that land rental and sale markets have consistently failed

to reallocate labor toward family farms. They detail several instances in Prussia, Latin

America and Southern Africa where the threat of land reform has led landlords to evict

tenants to reduce their exposure to expropriation (p. 2686-88).

Ireland, which did have a land reform, stands out as a very telling exception to the

pattern followed in much of rest of Europe. Landlords� property rights were under much

greater threat in Ireland because the vast majority of tenants were Catholic while landlords

were Protestants. Exactly as our analysis would predict, the extent of tenancy declined

markedly as the political conßict over land escalated, especially after the 1876 agrarian

crisis and leading up to independence in 1921. This period was marked by a number of

tenancy revolts, evictions, and escalating violence. Limited land reforms where attempted

in a vain effort to stave off more radical pressures from below (Swinnen, 2000). The area

of land under tenancy declined steadily from over 96 percent in 1879 to 70 percent in

1905, to 42 percent in 1910, to 25 percent in 1923. Following independence, the 1923

Irish Land Act bought out most of the remaining landlords.

The belief that tenancy experience leads to skill accumulation is widespread and has

formed the basis of economic theories of tenancy ladders which date back at least to

Spillman (1919), but empirical evidence is much harder to come by. Attack and Passell

20

(1994) summarize the debates on this matter for the United States. de Silva (1999)

provides recent econometric evidence of the signiÞcant difference in farming skills between

the landless, different types of tenants, and owners and reviews the literature. We

should note however that the skill accumulation connection is a convenient channel for

our model, and one that focuses attention on a plausible role for human capital, yet it is

in no way essential to our story. A very similar relationship between agrarian production

organization and the stability of property rights would emerge if landlords were concerned

only about tenants acquiring squatting rights and the possibility of a land-to-the-tiller type

land reform.

The following brief examples from Latin America and Asia further illustrate and

extend discussion of some of the issues raised by the model.

3.1 Colombia

Several scholars have pointed to the mechanism we have studied as an important deter-

minant of tenancy in Colombia. Both Palacios (1979) and Le Grand (1986) argued this in

the context of the Liberal agrarian reform of the early 1930�s. Zamosc (1986) documents

the evolving relationship between landlords and tenants in Colombia and its role in the

rise of a national peasant movement in the 1970s. The conßict over property rights is

illustrated by landlords� use of pasture-rent contracts to open up new frontier land in

the north-western regions of the country. Under this system peasants would clear forest

to open up new land in exchange for being allowed to grow rice, yuca, maize or other

food crops. At the end of a few years however tenants were required to sow pastures

and return the land to the landlord. The tenant was typically moved onto a new plot

of land that was fallow or needed to be cleared anew. Furthermore tenants were often

required to live in hamlets on the roadsides between haciendas rather than on the land

they farmed. These practices limited the tenant�s ability to establish possibly competing

property claims. In terms of our model, the use of short term shifting tenancies allows

landlord to extract some of gain to farming skill accumulation from tenancy (via higher

land rents), yet limits the political threat to their land.

The passage of national legislation 1968 granting potential rights to tenants brought

even this system to an abrupt end. According to Zamosc landlords expelled tenants �on

a massive scale, abolishing the customary patterns of access to land within a couple of

years (pp.78-79).� These expulsions were a major factor behind the strengthening of the

peasant movement and led to a wave of land invasions.

21

3.2 El Salvador

Land-scarce El Salvador has long been a place where conßicts over land have shaped

political outcomes, and vice-versa. Lauria-Santiago (1999) argues that, contrary to con-

ventional wisdom, until the early part of the twentieth century El Salvador was a country

with many smallholders, and relatively un-concentrated land holdings. Rapid population

growth, the closing of the agricultural frontier, the privatization of Indian lands, and the

expansion of coffee estates meant however that by the 1920�s a majority of the predomi-

nantly rural population was without ownership access to land. A large peasant uprising

in 1932 took place in regions where coffee estates drew labor from nearby farming com-

munities who felt their livelihoods and landholding status threatened by the downturn

of the depression. The extremely violent suppression of this uprising inaugurated a long

period of authoritarian-oligarchic rule supported by middle and large-scale landowners.20

The more secure property rights of landlords and changing relative prices led to an

initial �dramatic� rise in tenancy starting in the 1930�s (Lauria-Santiago, 1999, pp. 234-

235). The political equilibrium held for a few decades but over time came under pressure

as population growth, growing organization and militancy in the countryside and changing

political coalitions in the urban areas and military threatened to place land reform projects

back on the national agenda. Several authors point to the resulting �crisis in tenancy and

the role of mid-level landowners in expelling tenants and sharecroppers between the 1950�s

and 1970�s, triggering the political crisis that followed (Lauria-Santiago, 1999,p.238)�21

In 1972 a presidential candidate campaigned on a platform that openly called for

far-reaching agrarian reform. By most accounts the military stole this election and the

candidate was sent into exile. When a reformist Junta came to power in 1979 and

attempted to carry out land reform it met with staunch opposition and organized political

violence from landlords and the more conservative segments of the military. By the early

1980�s the country had plunged into a full-scale civil war.

Under strong US pressure, land reform was initiated in 1980 in the context of counter-

insurgency operations and to preempt further shift of support toward the leftist guerillas.

A proposed land-to-the-tiller reform led to another wave of tenant expulsions and increased

violence directed against peasants (Pelupessy, 1997). By the early 1990�s the civil war20Rural unions had already been banned in 1907 and a National Guard created a few years later with

the express purpose of protecting landlords� property rights and insuring labor tranquility. Nationalguard outposts were billeted to most large estates until the dismantling of the force in 1992.21Binswanger et al (1995) state that the number of house plots available to colono tenants decreased

from 55,000 to 17,000 in the 1961-70 period. See also Pelupessy (1997).

22

had come to a military stalemate. According to Paige (1999) and Wood (2000) resistance

to land reform and political democratization by the country�s elites lessened as the war

wore on as many landlords had abandoned land to squatters in the rural conßict zones

anyway, and been forced to diversify their wealth holdings out of agriculture. By the late

1980�s members of the economic elite with new interests in banking and urban activities

were able to persuade the ruling right-wing party to peace accords with the guerillas.

Our model points to a complementary explanation for this transition: the war, and

other structural transformations may have also reduced the expected value of land as

an asset in peasant hands, and hence lowered the pressure for, and the support of more

radical land reform. The large scale migration of the rural population to the cities and

to the United States and Canada transformed many former peasants into urban-dwellers

and created an important alternative avenue for economic advancement. In short, the

economy and the costs of war served to lower the potential gain to peasants from further

land asset redistribution just as much as it lowered the potential losses to landowners.

It also moved the median voter from the countryside to the cities. Each of these events

worked to lower the pressure for more radical asset redistributions, allowing both sides of

the conßict to move toward the bargaining table.

3.3 Chile

Despite having comparatively high levels of per capita income, urbanization, and edu-

cational attainment, Chile stands out for having retained a rural economy dominated

by traditional large landlord estates until well into the mid-twentieth century. Tenants

within large landlord estates were common, but these were typically long-term labor ser-

vice tenants (inquilinos) who were expected to provide family labor services in exchange

for access to small family plots (Gongora, 1960).

Sadoulet (1992) argues that labor service tenancy in Chile can be understood as an

interlinked contract to overcome credit and asymmetric information problems. Her analy-

sis is suggestive but this theory suggests an increase in the extent of tenancy over time as

tenants acquired the required wealth to lease in ever larger tracts of land. Chile�s system

of inquilinaje in fact evolved slowly and if anything, the number of sitting tenants declined

rather than expanded in the 20th century. According to de Janvry (1981) the number

of inquilinos peaked in the nineteen thirties and declined steadily thereafter as landlords

began to rely more heavily on hired wage labor. This trend was partly a response to

actual or anticipated legislation designed to protect inquilinos, and growing organization

23

in the countryside. Electoral reforms starting in 1958 expanded the size of the rural elec-

torate and reduced landlords� opportunities for vote manipulation by establishing secret

balloting. This soon led to new legislation lifting the ban on rural unionization and,

eventually far-reaching agrarian reforms in the 1965-1973 which led to the Þnal demise of

the large estates and the inquilinaje system.

Given the importance of mining and manufacturing and the size of the urban pop-

ulation in the twentieth century, Chile appears to follow the pattern of Uruguay and

Argentina in having kept land reform off the national agenda until well into the 1960�s.

Why then did land reform emerge with so much force when it did? Although the median

voter was not likely a peasant, traditional rural-urban cleavages seem to have temporarily

given way to rich-poor cleavages in shaping the pattern of electoral competition. By

1970 all three principal presidential candidates from left to right had made land reform a

central plank of their programs. The reason appears to be that political parties were ag-

gressively competing for the 25 percent of the electorate that had been released following

the 1958 electoral reforms (Baland and Robinson, 2000). Kaufmann (1970) argues how-

ever that such a transformation in the dynamics of electoral competition were destined

to be short lived as urban workers�s welfare soon began to be affected by the rising prices

brought about by agricultural disruption and by the increasing claims that the reform

programs placed on scarce government revenues. A military coup in 1973 put an abrupt

end to the redistribution programs and Þrmly re-established property rights security. By

then however the large landlord estates had been destroyed. Land rental markets in Chile

today are very active by Latin American standards.

3.4 Korea

Jeon and Kim (2000) offer an interesting account of the process of land redistribution

in Korea that illustrates the important political role of tenants. Tenancy under the

Japanese colonial administration 1919-45 had been high: in 1939 over 58 percent of farm-

land was under tenancy and 56 percent of farmer households were tenants. Although

tenant protests demanding lower rents were not uncommon under Japanese colonial rule,

the Japanese military presence strictly enforced landlord�s property rights. Landlord po-

litical power was very seriously and quickly eroded however when the Japanese were forced

to abandon the Korean peninsula and Korea came under the US military administration

in August 1945 however.

Both the US military administration and the Þrst democratically elected administra-

24

tion in 1948 signaled a Þrm willingness to implement land reform legislation. The North

Korean threat helped to remove the remaining political impediments and land reform

legislation was passed in 1950. Although the reforms had the appearance of being exter-

nally imposed, Jeon and Kim point out that the 1948 land reform should be understood

as �an endogenously determined governmental policy consistent with the intuition of the

median voter theorem ... [as] tenants represented the largest portion of the population

(pp. 257-258).� In fact, the effects of the anticipation of land reform were felt as soon

as the Japanese withdrew, years before Þnal reforms were enacted in 1950. During this

period the tenancy market �broke down� as tenants collectively refused to pay rents. The

eventuality of reform became so certain however that 60 percent of landlords � mostly the

larger ones � sold their land to tenants via the market at reduced prices before 1950. In

fact more than twice as much land was sold by landlords in anticipation of the reform than

was ever transferred directly via the land reform law (p. 255, Table 1). Jeon and Kim

conclude their analysis by arguing econometrically that land reform increased agricultural

productivity by enhancing economic incentives.22

4 Conclusion

The modern theory of agrarian organization has studied how the economic environment

determines organizational form under the assumption of exogenous property rights to

land. The political economy literature has modelled the endogenous determination of

property rights and the distribution of land ownership. In this paper we have argued that

the form of agrarian organization may also be inßuenced by the anticipation of property

rights challenges. In particular, we argued that landowners may have an incentive to limit

the extent of tenancy to reduce the expected extent of land redistribution. Despite the

economic advantages that tenancy embodies, by giving tenants de facto property rights

or increasing their skills, it also raises the desired extent of land reforms via the political

process.

Though the economic environment undoubtedly plays an important role in agrarian

organization, we argued that our theory provides an important key to why there seems to

be so little tenancy in situations where agrarian reform is a salient political issue, particu-

larly in Latin America. Our prediction that initial asset inequality can slow growth via its

impact on private agents� decisions to alter production organization to protect property22They argue that since much of the land under tenancy had previously been sharecropped, the transfer

of ownership rights removed a classic Marshallian inefficiency.

25

rights is supported by recent evidence and accounts, including Keefer and Knack�s (2000)

cross-country econometric Þndings that inequality reduces growth primarily via its effect

on property rights security rather than through several of the other channels hypothesized

in the literature.

There is an important implication of our model. Interestingly, the very success of

agrarian reform seems intimately tied to pre-reform agrarian organization. Because the

extent of tenancy was high, land reform beneÞciaries in East Asia had already acquired

farming skills and management experience which meant that production was not sub-

stantially disrupted by the transfer of property rights. Land reform in these countries

is deemed to have been highly successful both at raising agricultural productivity and

at releasing labor into the newly expanding sectors that formed the basis of East Asia�s

economic miracle (Dore, 1959; Tuma, 1965).23 In most countries of Latin America where

land reform has taken place land was often transferred to landless farm laborers or labor

service tenants with much less practical experience as independent producers and with low

educational attainment. Not surprisingly, land reform was initially far more disruptive

to production, and has ultimately been less successful at raising agricultural productivity

and incomes. Our theory thus helps to account for, not just the relative incidence of dif-

ferent forms of agrarian organization, but also the effects of a transfer of property rights

on the efficiency of the rural economy.24

23Both Banerjee, Gertler and Ghatak (1996) and Besley and Burgess (2000) present evidence of positiveimpacts of tenancy and agrarian reforms in West Bengal and India, but these instances fall far short ofthe mass redistributive changes seen in East Asia or attempted in Latin America.24Though we did not model this phenomenon formally it would be easy to extend the model to capture

it. Intuitively, land reform induces costs because individuals have to develop speciÞc productive skills andundertake tasks which are new to them. When agents with little experience are given land output fallswhile they develop such skills. However, the extent to which they already have such skills is a functionof the extent of tenancy in the Þrst period. Thus tenancy mitigates the �disorganization� caused by landredistribution. To capture this we could allow s to be decreasing in α hence, s(t, h,α) and sα < 0 but withsαt > 0. Such a formulation adds an extra cost to land redistribution so the details of the expressionschange but the interesting new result would be that the initial extent of tenancy could be the differencebetween land reforms where agricultural output rise or falls after reform.

26

Bibliography

Acemoglu, Daron and James A. Robinson (1999) �A Theory of Political Tran-

sitions, CEPR Discussion Paper #2277, forthcoming in the American Economic Review.

Aghion, Philippe and Patrick Bolton (1990) �Government Domestic Debt and

the Risk of a Default: A Political-Economic Model of a Strategic Role of Debt,� in Rudi-

ger Dornbusch and Mario Draghi eds. Public Debt Management, Cambridge, Cambridge

University Press.

Alesina, Alberto (1988) �Credibility and Policy Convergence in a Two-Party Sys-

tem with Rational Voters,� American Economic Review, 78, 796-805.

Alesina, Alberto and Guido Tabellini (1990) �A Positive Theory of Fiscal

DeÞcits and Government Debt,� Review of Economic Studies, 57, 403-414.

Baland, Jean-Marie and James A. Robinson (2000) �Land and Power,� Un-

published, Department of Political Science, UC Berkeley.

Banerjee, Abhijit and Andrew F. Newman (1993) �Occupational Choice and

the Process of Development,� Journal of Political Economy, 101(2): 274-98.

Banerjee, Abhijit, Paul Gertler and Maitreesh Ghatak (1996) �Empower-

ment and Efficiency: The Economics of Tenancy Reform,� Unpublished, Massachusetts

Institute of Technology.

Bardhan, Pranab (1989) The Economic Theory of Agrarian Institutions, Oxford;

Oxford University Press.

Bardhan, Pranab and Christopher Udry (1999) Development Microeconomics,

New York, Oxford University Press.

Basu, Kaushik (1997)Analytic Development Economics: The Less Developed Econ-

omy Revisited, Cambridge, MA, MIT Press.

Berry, R. Albert (1972) �Farm Size Distribution, Income Distribution, and the

Efficiency of Agricultural Production: Colombia,� American Economic Review, 62(2):

403-08.

Besley, Timothy (1995) �Property Rights and Investment Incentives: Theory and

Evidence from Ghana,� Journal of Political Economy, 103, 903-937.

Besley, Timothy and Robin Burgess (2000) �Land Reform, Poverty Reduction

and Growth: Evidence from India,� Quarterly Journal of Economics, 115, 389-430.

Besley, Timothy and Stephen Coate (1997) �An Economic Model of Represen-

tative Democracy,� Quarterly Journal of Economics, 112: 85-114.

27

Besley, Timothy and Stephen Coate (1998) �Sources of Inefficiency in a Repre-

sentative Democracy: A Dynamic Analysis,� American Economic Review, 88, 139-156.

Binswanger, Hans P., Klaus Deininger, and Gershon Feder (1995) �Power,

Distortions, Revolt and Reform in Agricultural Land Relations,� in Jere Behrman and

T. N. Srinivasan eds, Handbook of Development Economics, Volume IIIB, Amsterdam,

North Holland.

Bourguignon, Francois and Thierry Verdier (2000) �Oligarchy, Democarcy,

Inequality and Growth,� Journal of Development Economics, 62, 285-213.

Cheung, Stephen N. (1969) A Theory of Share Tenancy, Chicago, University of

Chicago Press.

Collier, D., R. Collier and R. Berins (1991) Shaping the Political Arena, Prince-

ton University Press.

Conning, Jonathan (2000). �Latifundia Economics,� Williams College, Depart-

ment of Economics, working paper.

de Janvry, Alain (1981) The Agrarian Question and Reformism in Latin America,

Baltimore, Johns Hopkins University Press.

de Janvry, Alain and Elisabeth Sadoulet (1989) �A Study in Resistance to Insti-

tutional Change: The Lost Game of Latin American Land Reform.� World Development,

17: 1397-1407.

Deinenger, Klaus and Pedro Olinto (2000). �Asset Distribution, Inequality, and

Growth,� World Bank working paper, Washington, DC.

Dixit, Avinash K. and John B. Londregan (1996) �The Determinants of Success

of Special Interests in Redistributive Politics,� Journal of Politics, 58, 1132-1155.

Dore, Ronald P. (1959) Land Reform in Japan, Oxford; Oxford University Press.

Eswaran, Mukesh and Ashok Kotwal (1984) �Access to Capital and Agrarian

Production Organization,� Economic Journal, 96: 482-498.

Fei, John C.H., Gustav Ranis and Shirley W.Y. Kuo (1979) Growth with

Equity: The Taiwan Case, New York, Oxford University Press.

Góngora, M. (1960). Origen de los inquilinos de Chile central. Santiago, Universi-

dad de Chile, Seminario de Historia Colonial.

Grossman, Herschel I. (1994) �Production, Appropriation, and Land Reform,�

American Economic Review, 84(3): 705-12.

Hayami, Yujiro and Keijiro Otsuka (1993) The Economics of Contract Choice:

An Agrarian Perspective, Oxford; Oxford University Press.

28

Horowitz, Andrew W. (1993) �Time Paths of Land Reform: A Theoretical Model

of Reform Dynamics,� American Economic Review, 83(4): 1003-1010.

Jeon, Y.-D. and Y.-Y. Kim (2000). �Land Reform, Income Distribution, and

Agricultural Production in Korea.� Economic Development and Cultural Change: 253-

268.

Kaufman, Robert R. (1972). The politics of land reform in Chile, 1950-1970;

public policy, political institutions, and social change. Cambridge� Harvard University

Press.

Keefer, Philip and Stephen Knack (2000). �Polarization, Politics, and Property

Rights: Links between Inequality and Growth,� World Bank working paper.

Kevane, Michael (1996) �Agrarian Structure and Agricultural Practice: Typology

and Application to Western Sudan,� American Journal of Agricultural Economics, 78(1):

236-45.

Lapp, Nancy D. (1997) Landing Votes: Expansion of the Suffrage and Land Re-

form in Latin America, Unpublished Ph.D. Dissertation, Department of Political Science,

UCLA.

LeGrand, Catherine (1986) Frontier expansion and Peasant Protest in Colombia,

1850-1936, Albuquerque, University of New Mexico Press.

Legros, Patrick and Andrew F. Newman (1996) �Wealth Effects, Distribution,

and the Theory of Organization,� Journal of Economic Theory, 70(2): 312-41.

Lindbeck, Assar and Jurgen Weibull (1987) �Balanced Budget Redistribution

as the Outcome of Political Competition,� Public Choice, 52, 273-297.

Moene, Karl Ove (1992) �Poverty and Landownership,� American Economic Re-

view, 82(1): 52-64.

O�Donnell, Guillermo (1979) �State and Alliances in Argentina, 1956-1976,� Jour-

nal of Development Studies, 15, 3-33.

Palacios, Marco (1979) El Café en Colombia, 1850-1970: Una Historia Económica,

Social y Politica, Editorial Presencia, Bogotá, Colombia.

Paige, Jeffrey M. (1996) Land Reform and Agrarian Revolution in El Salvador:

Comment on Seligson and Diskin. Latin American Research Review 31(2): 127-139.

Pelupessy,Wim (1997) The limits of economic reform in El Salvador, New York,

St. Martin�s Press.

Persson, Torsten and Lars O. Svensson (1989) �Why would a Stubborn Con-

servative Government run a DeÞcit? Policymaking with time Inconsistent Preferences,�

Quarterly Journal of Economics, 104, 325-345.

29

Persson, Torsten and Guido Tabellini (2000) Political Economics: Explaining

Economic Policy, Cambridge, MIT Press.

Prosterman, Roy L. and Jeffrey M. Riedinger (1987) Land reform and demo-

cratic development, Baltimore, Johns Hopkins University Press.

Robinson, James A. (1998) �When is a State Predatory?� Unpublished, Depart-

ment of Political Science, UC Berkeley.

Scott, James C. (1985). Weapons of the weak : everyday forms of peasant resis-

tance. New Haven, Yale University Press.

Sokoloff, Kenneth L. and Stanley L. Engerman (2000). �Institutions, Fac-

tor Endowments, and Paths of Development in the New World.� Journal of Economic

Perspectives 14(3).

Swinnen, J. (2000). The Political Economy of Institutional Change: A historical

perspective on land tenure in Western Europe. Policy Research Group, Department of

Agricultural and Environmental Economics, Katholicke Universiteit Leuven.

Stiglitz, Joseph E. (1974) �Incentives and Risk Sharing in Sharecropping,� Review

of Economic Studies, 41: 397-426.

Tuma, E. H. (1965). Twenty-six centuries of agrarian reform; a comparative analy-

sis. Berkeley, University of California Press.

Taylor, Alan M. (1997) �Latifundia as Malefactor in Economic Development?

Scale, Tenancy, and Agriculture on the Pampas, 1880-1914,� NBERWorking Paper Series

on Historical Factors in Long Run Growth # 96.

Wilkie, James (1996) Statistical Abstract of Latin America.

Williams, Richard G. (1986). Export agriculture and the crisis in Central America.

Chapel Hill, University of North Carolina Press.

Wood, Elisabeth J. (2000). Forging Democracy from Below: Insurgent Transitions

in South Africa and El Salvador, Cambridge University Press.

Zamosc, Leon (1986) The Agrarian Question and the Peasant Movement in Colom-

bia : Struggles of the National Peasant Association, 1967-1981, Cambridge, Cambridge

University Press.

30

Table 1: Distribution of farmland by land tenure status, 1970 World Census of Agriculture

Asia Africa Latin

America Europe North

America World Number of Countries 10 4 15 12 2 46Number of Farms (millions) 93.3 3.5 8.6 11.9 3.1 120.4Avg. operational farm size (ha) 2.3 0.5 46.5 7.6 161.2 10.0Percent of farmland under:

Pure owner Cultivation 84.0 9.2 80.4 58.9 36.6 61.1Pure Tenancy 5.9 3.0 6.2 12.5 11.9 9.0Owner-cum-Tenancy 10.1 29.1 5.6 28.5 51.5 27.2Communal or other 0.0 58.7 7.8 0.1 0.0 2.7

Source: Adapted from Table 1.1. in Hayami and Otsuka (1993) The Economics of Contract Choice, Oxford.

Notes: Farmland under owner-cum-tenancy includes both owned and leased land. Asia: Bahrain, India, Indonesia, Jordan, Korea, Kuwait, Pakistan, Philippines, Saudi Arabia, Singapore. Africa: Cameroon, Reunion, Swaziland; Latin America: Costa Rica, Dominican Republic, El Salvador, Guadeloupe, Honduras, Panama, Puerto Rico, St. Lucia, Virgin Islands, Brazil, Colombia, Peru, Surinam, Uruguay, Venezuela; Europe: Austria, Belgium, France, West Germany, Italy, Malta, Netherlands, Norway, Poland, Portugal, Sweden, UK; North America: Canada, USA.

Table 2: Land Ginis and area cultivated by tenure status in selected countries

YearLand Gini

Pure Tenant

Tenancy Totala

ASIA Bangladesh 1976 0.42 - 20.9 India 1970 0.62 2.4 8.5 Indonesia 1973 0.56 2.1 23.6 Philippines 1971 0.51 21.4 32.8 Thailand 1978 0.45 6 15.5 Taiwan 1939 - 56.3

1959 - 14.4 Korea 1939 - - 58.4 1955 0.34 0.5 Japan 1941 0.42 45.8 EUROPE Belgium 1880 - 64.0 1990 - 67.0 Ireland 1870 - 96.0 1990 0.62 6.0 France 1880 - 40.0 1990 0.54 57.0 UNITED STATES 1969 40.0 1997 0.73 49.0 LATIN AMERICA Argentina 1960 0.79 14.6 Brazil 1970 0.84 6.1 10.2 Bolivia 1950 0.94 7.5 Costa Rica 1973 0.82 1.2 9 Chile 1965 0.94 14.2 24.4 Colombia 1960 0.86 5.3 11.5 El Salvador 1961 0.81 7.8 Nicaragua 1963 - 2.6 Peru 1961 0.91 4.5 13.6 Uruguay 1970 0.82 19.1 46.3 Venezuela 1961 0.91 4.5 2.4

Notes: Non-italics numbers in the Tenancy Total column indicate the fraction of cultivated land under pure tenancy or owner-cum-tenancy (i.e. the data do not distinguish between leased and owned land in the owner-cum-tenancy subcategory). Italicized figures report only the fraction of cultivated land actually under tenancy (i.e. not counting owned land in the owner-cum-tenant subcategory).

Sources: Asian country data except for Taiwan and Korea from Table 1.2 in Hayami and Otsuka (1993); Taiwan: Fei, Ranis & Kuo (1979). Korea: Jeon and Kim (2000), Table A1; Japan: Tuma (1965), Table 23. Europe: Swinnen (2000), various tables. Latin America and 1969 data for United States: Wilkie, J. (ed.) (1996), Tables 200, 201, 206. United States: 1997 Census of Agriculture. Land Ginis for USA, Ireland, and France from Deininger and Olinto (2000), Table 2.